A Familiar Tale
I recently heard a familiar tale from the CIO of a Fortune company: “We don’t have a repeatable or accurate cost allocation model that can be used to reflect the true cost of computing services consumption back to our lines of business. Quantifying a workload’s owner/size/scope and forecasting the growth profile and scale of our shared hosting and cloud environments has proved to be very challenging.”
What this CIO and his organization need are what I am calling a “smart cloud.”
In this post I will illuminate the idea of a smarter cloud in ways that I hope you can action. I will answer several common questions I get from clients who are looking for ways to be smarter about managing their spend on the cloud:
- How to perform an apples-to-apples comparison based on an economic measure of IT across internal and external infrastructures and suppliers
- How to show the clear economic benefits of using cloud hosting—particularly when the economic benefits have been murky or negative
- How to baseline, benchmark, and forecast IT usage by measuring the total cost of consumption, vs. outdated allocation models
So, let’s get to it.
Why Your Cloud Should be Smarter
What do I mean by a smart cloud? What are its characteristics and why should you care?
Well, you should care for one very important reason alone: by deploying a smarter shared application hosting and / or cloud environment, you can realize as much as a 50% ongoing CapEx and OpEx reduction. Really, I kid you not—I have helped clients achieve these kinds of outcomes. I liken the provisioning of shared application hosting services to an organization, regardless of the deployment model, to feeding goldfish. The more you feed them, the more they consume. Eventually they die from overeating.
Put into terms that are relevant to IT compute services: if you don’t have a way to effectively manage and forecast demand, you are doomed to a life of over-provisioning and runaway costs, e.g., overeating.
How to Make Your Cloud Smarter
Whether you are managing plain old shared services, public, hybrid, private cloud services (referred to here collectively as the “cloud”), or all of the above, you need a way to accurately measure and allocate consumption costs to business users based on a common metric. Of equal importance, and what is often overlooked, is that you need a way to demonstrate efficiency and waste to your customers. In other words, don’t let them act like goldfish!
Okay, so I had you at hello. Making your cloud smarter means first addressing the issues of cost transparency, effective demand management, and efficient capacity utilization. If wasted capacity cannot easily be measured or monetized, you will most certainly experience unnecessary infrastructure growth.
Smart Cloud Capabilities
While the devil is most certainly in the details, fundamentally you need three sets of capabilities to make your cloud smarter:
- Real-time consumption monitoring tied to the cost of computing with the added capability to measure and report on waste.
- Ability to enable workload- and business service-level compute cost allocation.
- Ability to determine the least-cost placement path for different classes of workloads to optimize their ongoing spending profiles.
Seven Steps to a Smarter Cloud
Assuming you are early in your journey to activating your “smart cloud,” here are seven steps you can follow as a structured approach:
- Ascertain your current organizational and compute consumption/resource allocation mapping by identifying project cycles within organizational divisions, which in turn impose demand cycles on current compute resource allocations, e.g., CPU in Mhz, MEM in MB, STORAGE in GB, DISK I/O in KBps, LAN I/O in Kbps, and WAN I/O in Kbps.
- Derive your organization structure, mapping business-line decision-makers to IT support infrastructure to business service application workload owners.
- Gain organizational access to those who have operational knowledge of all related compute demand that is currently in service to all business lines.
- Derive business services and component workload matrices related to service tiers, workload class, and minimum / target SLA / QoE (Quality of Experience) for spin-up / spin-down within required performance windows for each.
- Document hardware and workload orchestrations detailing assignment of workloads to infrastructure.
- Quantify current contracted and amortized cost for business service member workloads for supporting infrastructure and for computing resource consumption over appropriate time windows needed to reflect demand peaks (net cost of computing, or NCC).
- Map the information derived in steps 1-6 to create an organization-wide compute cost and waste allocation model for business service member application workloads.
Where to Next?
Once you have gone through the seven-step process, you should at minimum have actionable insight and analysis into:
- Your organization’s inventory of business services and contributing workloads captured and considered for efficiency and waste trade-off by Quality of Experience (QoE).
- Current compute spend by business service with the ability to drill into contributing workloads.
- Workloads, holistically, with respect to peaks that “cost justify” existing contract allocation. Contract services may require adjustment where business service QoE is impacted by peaks’ encroachment on contracted thresholds.
Armed with these insights, you will be in a great position to show the true cost of business service / application workload consumption and, importantly, wasted cost back to your customers.
You will also be in prime position to exploit alternative suppliers that could optimize NCC or optimize increased QoE / QoS for key business service workloads.
In case you want to take advantage of the extensive work we have done in this area, you can download our RampRate Digital Services Efficiency Dashboard for ITBM data sheet and learn all about RampRate’s powerful infrastructure efficiency dashboard, which enhances your investment in ITBM tools, supported by RampRate SPY INDEX process and SPY™ data services.